Detection of foliar diseases using image processing techniques

ABSTRACT This paper presents the development of a methodology to detect the percentage of affected area of Phytophthora infestans disease in tomato plants, using digital image processing techniques to extract the regions of interest with color analysis, where the YIQ and TSL models for the detection of the disease. The method consists of solving one of the most common problems in images that is segmentation, in this case the background and the disease with the Plant Village database, which was captured under uncontrolled lighting conditions. In the experiments conducted, it is observed that our method achieved a performance of 98.60% for the detection of healthy pixels and 98.17% in detection of sick pixels. This process was subjected to comparison against other alternatives of the state of the art like K-means with HSV and LAB, showing a referred error regarding the leaf size of 4.32 ± 5.44% in the detection of the disease and a computational time of 0.03 ± 0.01 [s] in comparison with the other procedures, in addition, this methodology was implemented to detect the foliar diseases black Sigatoka and yellow Sigatoka in banana leaves obtaining satisfactory results.

摘要 本文提出了一种检测番茄植株晚疫病（Phytophthora infestans）侵害面积占比的方法体系，通过数字图像处理技术结合颜色分析提取感兴趣区域，并采用YIQ与TSL颜色模型实现病害检测。该方法旨在解决图像领域最常见的问题之一——图像分割，本次实验依托植物村落（Plant Village）数据库完成背景与病害区域的分割，该数据集采集于非受控光照环境下。实验结果表明，本方法的健康像素检测准确率达98.60%，病斑像素检测准确率达98.17%。将本方法与当前前沿的其他替代方案（如结合HSV与LAB颜色空间的K-means算法）进行对比，本方法在病害检测中相对于叶片尺寸的参考误差为4.32±5.44%，计算耗时为0.03±0.01[s]。此外，本方法还被拓展应用于香蕉叶片的黑条叶斑病（black Sigatoka）与黄条叶斑病（yellow Sigatoka）检测，均取得了令人满意的实验结果。